What do you still check manually before shipping an AI-built site?

AI can build and ship a website quickly, but “it runs” is not the same as “it’s ready.”

Before you publish, which checks still slow you down most: security, SEO, AI visibility, accessibility, performance, or integrations?

I’m building Sitelemetry to scan those areas, turn findings into evidence-backed tasks, and pass them to Codex or Claude through MCP so the agent can fix and re-test. It can also audit localhost through the Local Agent before a domain exists.

I’d love to learn from your workflow:

• What stack are you using?

• Which check do you usually miss or postpone?

• Would you trust an agent to apply low-risk fixes automatically, or do you want approval for every change?

I’ll use the answers here to prioritize what we improve after launch.